iii
Self-hosted distributed runtime for multi-language workers and AI agents under one protocol.
Pick iii if you're an engineering team that wants one self-hosted runtime for multi-language workers and AI agents without gluing five SaaS services together.
Skip it if you want a managed, no-ops agent platform or a polished prebuilt agent UI rather than a low-level distributed runtime.
iii (pronounced "three eye") is an open-source distributed software platform that unifies workers, functions, and AI agents across Python, Rust, Node.js, and Go behind a single protocol. Instead of stitching together separate queues, state stores, vector indexes, and observability tooling, iii ships them as built-in primitives, with a unified console that surfaces workers, functions, triggers, streams, queues, traces, and logs in real time. AI agents are treated as ordinary workers, so planning, tool-running, and synthesis steps run on the same engine as the rest of the backend.
The platform is aimed at engineering teams building distributed systems or agentic applications who are tired of the quadratic integration cost of gluing together five SaaS dependencies. It bundles durable orchestration, OpenTelemetry tracing, WebSocket streaming, HTTP endpoints, event triggers, cron scheduling, queuing with backpressure, and vector storage. Install is a single curl-piped shell script and it runs locally (127.0.0.1) by default, so it sits closer to a self-hosted Temporal/Inngest hybrid than a managed agent cloud.
The project lives on GitHub (iii-hq/iii) with an active Discord. Pricing isn't published — the GitHub-first, self-hosted posture suggests the open-source core is free with a paid commercial tier likely to follow. Model integrations aren't dictated by the platform; you bring your own LLM in whichever language your worker is written in.
iii is an ambitious attempt to collapse the agent stack and the distributed-systems stack into one engine, and the multi-language worker protocol is the genuinely interesting bit. It's early — docs and pricing are thin — but worth a look for teams already self-hosting Temporal, Inngest, or a homegrown worker mesh.
— The AI Tool Bible editorial team
Pros
- ✅ One protocol for Python, Rust, Node.js, and Go workers
- ✅ Bundles queues, state, vectors, and OpenTelemetry tracing out of the box
- ✅ Treats AI agents as ordinary workers — no separate agent runtime
- ✅ Open source with single-command install
Cons
- ⚠️ Pricing and managed-cloud story not yet published
- ⚠️ Early-stage project with thin public documentation
- ⚠️ Self-host operational burden vs hosted competitors
Use cases
Explore related
Compare with similar tools
All in Agents →LangGraph
FeaturedStateful, graph-based agent orchestration from LangChain.
CrewAI
FeaturedPython framework for multi-agent orchestration.
Claude Agent SDK
Anthropic's official SDK for building autonomous Claude agents.
Manus
Generalist agent for research, code, and web tasks.
Devin
Cognition Labs' "autonomous software engineer" agent.
AutoGPT
Open-source platform for building autonomous AI agents.